Consensus-based data replication protocol for distributed cloud

Abstract

Data availability ensures efficient data accessibility by the readers anytime and from anywhere. It can be addressed by creating multiple copies of each data file and storing them on well-distributed distinct servers. The more the number of copies, the more is the availability. Considering a distributed cloud scenario with multiple data copies, a file update operation may be performed at any server containing a copy of the data file. Allowing parallel file updates by different users on various servers may incur inconsistent views of the data file among readers. A data replication protocol ensures that a file will remain consistent throughout the network. The existing data replication protocols did not explicitly address the server confidence about when the updated file version will be ready for read. In this work, we propose a consensus-based file replication protocol considering the message passing model that addresses the server confidence issue of the existing protocols. In the proposed protocol, the updated data file will be immediately accessible to the readers without any ambiguity after consensus is made. The proposed protocol is analyzed and compared with the similar protocols. The protocol is implemented, and the experimental results are verified with the analytical results.